The rapid growth of generative AI makes technology safety a top priority. On February 25, 2026, OpenAI released a new research report. The report is titled "Disrupting Malicious Uses of AI." It summarizes two years of insights into AI abuse. It also reveals several real-world case studies. Today, threat activities rarely stay on a single platform. Instead, actors combine AI with social media and fake websites. They create complex collaboration chains. These actors often switch between different AI models to suit specific tasks. This cross-platform evolution presents a new challenge to digital security.
AI is now deeply integrated into scams and influence operations. Criminal groups have standardized their process into three stages: contact, building trust, and theft. For example, some scams use AI to play mentors or romantic partners. They lure victims into paying high fees. Some even use AI to create fake legal documents and licenses. They pose as "scam recovery services" to target previous victims again. Furthermore, some state-linked actors use AI for mass propaganda. They generate geopolitical articles while posing as experts. They also target political figures and dissidents. Their tactics include "squatting" on fake accounts to drown out real information. They also use AI to trigger mass reporting mechanisms. This is done to manipulate public perception.
The report shares proactive detection and collaboration mechanisms. OpenAI continuously strengthens internal safeguards. As a result, models now explicitly refuse requests related to malicious operations. OpenAI also tracks specific behavioral patterns. This allows them to accurately identify and ban violating accounts. Major platforms and law enforcement agencies now share intelligence. This extends defense across the entire industry. Identifying threats is no longer just about detecting AI-generated content. Users should observe account authenticity and interaction patterns. Be vigilant toward missing professional licenses or redirects to private messaging apps. Stay alert for recruitment messages that seem too good to be true. By staying cautious, we can enjoy AI's benefits while maintaining safety and truth.